Writing · Beginner
Template 11: Result sentences, figure legends and table notes
Use these patterns to expose the estimate and design. Replace every bracket with a verified value or description. Remove patterns that do not fit your analysis.
4 min read · Oct 26, 2026
Key points
- Adapt the prompts to the actual study.
- Complete every statement from verified records.
- Follow the target journal and institutional requirements.
Contents
Use these patterns to expose the estimate and design. Replace every bracket with a verified value or description. Remove patterns that do not fit your analysis.
Use these patterns to expose the estimate and design. Replace every bracket with a verified value or description. Remove patterns that do not fit your analysis.
Continuous comparison
“Among [n and independent unit], [outcome] was [summary with defined spread] in [A] and [summary] in [B]. The estimated [A minus B or other defined contrast] was [value and units], with a [level and method] confidence interval from [lower] to [upper]. [Test statistic, degrees of freedom and p value if relevant.]”
Paired comparison
“Each [unit] contributed measurements under [conditions]. The mean within unit change, defined as [direction], was [estimate and units], with [interval]. The analysis included [n] complete pairs; [missingness and handling].”
Categorical result
“[Count] of [denominator] [units] met [defined criterion], corresponding to [percentage]. The estimated [risk difference, ratio or other effect] comparing [groups] was [estimate and interval].”
Regression result
“A [one unit or other defined] difference in [predictor] was associated with [coefficient and units] in [outcome], with [interval], conditional on [specified model covariates]. [State design limitations on causal interpretation.]”
Uncertain finding
“The estimate was [value], with an interval from [lower] to [upper]. This range includes [scientifically relevant possibilities], so the analysis [state what it can and cannot distinguish].”
Do not automatically translate p > 0.05 into “no effect”. If equivalence is the question, use an appropriate prespecified margin and analysis.
Figure legend
Title: [What the figure shows]
Panels: [A: description; B: description]
Observations: [What each point, line, bar or image represents]
Sample: [Independent n and any nested measurement count]
Summary: [Mean, median, model estimate or other quantity]
Uncertainty: [SD, SE, interval level and calculation]
Encoding: [Colours, shapes, lines, reference values and abbreviations]
Analysis: [Comparison, test or model and multiplicity where relevant]
Image details: [Scale bar, processing, crops or composites if applicable]
Provenance: [Original data, clearly labelled synthetic data or conceptual illustration]
Alt text: [Purpose and main pattern, with key information accessible without colour]
Table note
“Values are [summary definition]. Denominators refer to [definition] and vary by row where shown. [Missing symbol] indicates [meaning]. The reference category is [category]. Intervals are [definition]. P values are [raw or adjusted] using [method] across [family]. [Explain abbreviations and units.]”
Final cross check
Open the analysis output, figure export and manuscript together. Verify direction, values, units, sample sizes, interval method and category names. A later analysis change requires updating all three.
Tutorial: separate the display, the result and the interpretation
Read Article 10, Article 19 and Article 25. A legend explains how to read a display. A Results sentence identifies the finding that matters. The Discussion explains its possible meaning and limits. Avoid putting unsupported explanations into a legend just because the figure looks persuasive.
Synthetic worked result: “The twelve constructed paired cases had a mean reduction of 3.58 mm, with a 95% confidence interval from 2.95 to 4.22 mm.” A corresponding legend specifies A minus B, millimetres, one line per paired case and the t based interval. It states that the data are synthetic. The paragraph need not repeat all twelve rows, but it must preserve the direction and denominator.
For social science or qualitative results: State the denominator for percentages and the level at which observations are independent. Introduce quotations with enough context to support interpretation while protecting identity. Explain how the quoted material supports the analytical theme; a vivid quotation alone does not establish prevalence.
Completion check: Print or view the figure separately from the manuscript. Can a reader identify the outcome, units, observations, sample size, symbols and uncertainty measure? Inspect the same figure in grayscale or using a colour vision simulation when distinctions rely on colour.